The CEO of OpenAI, Sam Altman, has revealed that the company has had to "temporarily" throttle image generation requests due to the overwhelming popularity of ChatGPT's image creation feature. In a social media post, Altman stated, "We're thrilled to see people enjoying ChatGPT's image capabilities, but our GPUs are struggling to keep up." He did not provide specifics on the extent of the restrictions but expressed hope that this measure would not last long, as OpenAI is working to enhance efficiency to handle the surge in demand.

Previously, in response to skyrocketing demand, OpenAI delayed access to the built-in image generator for free-tier ChatGPT users. However, this step was evidently insufficient to ease the pressure on OpenAI’s infrastructure. Altman noted that free users would "soon" be able to generate up to three images per day.

OpenAI leverages its GPT-4o model to boost the performance of its image generation tool. The updated model produces more realistic images and has made advancements in areas like text rendering, which were previously challenging. Nevertheless, this latest throttling serves as a reminder of the immense computational resources (and energy) required to generate images within ChatGPT.